Re: [PATCH v1 2/4] docs: stable-kernel-rules: mention "no semi-automatic backport"

[Greg Kroah-Hartman]

On Thu, Apr 11, 2024 at 11:57:04AM +0200, Thorsten Leemhuis wrote:
> On 11.04.24 11:19, Geert Uytterhoeven wrote:
> > On Thu, Apr 11, 2024 at 11:13 AM Greg Kroah-Hartman
> > <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> >> On Thu, Apr 11, 2024 at 09:50:24AM +0200, Thorsten Leemhuis wrote:
> >>> On 11.04.24 09:40, Greg Kroah-Hartman wrote:
> >>>> On Thu, Apr 11, 2024 at 08:59:39AM +0200, Thorsten Leemhuis wrote:
> >>>>> On 11.04.24 07:29, Greg Kroah-Hartman wrote:
> >>>>>> On Thu, Apr 11, 2024 at 07:25:04AM +0200, Thorsten Leemhuis wrote:
> >>>>>>> Some developers deliberately steer clear of 'Fixes:' tags to prevent
> >>>>>>> changes from being backported semi-automatically by the stable team.
> >>>>>>> That somewhat undermines the reason for the existence of the Fixes: tag,
> >>>>>>> hence point out there is an alternative to reach the same effect.
> >>>> [...]
> >>>>>> I do not understand, why are you saying "cc: stable" here if you do NOT
> >>>>>> want it backported?
> >>>>> Because the only alternative the developers have to make the stable team
> >>>>> not pick a single patch[1] is to deliberately omit a Fixes: tag even if
> >>>>> the patch normally should have one. Like it was done here:
> >>>>> https://lore.kernel.org/all/cover.1712226175.git.antony.antony@xxxxxxxxxxx/
> >>>> That feels odd, but ok I now see the need for this for some minor set of
> >>>> changes (i.e. this has rarely come up in the past 15+ years)
> >>>>
> >>>> [...]
> >>>>> E.g. 'ignore for the AUTOSEL and the "Fixes tag only" tools'. That was
> >>>>> the best term I came up with.
> >>>>
> >>>> Thinking about it more, I think we need to be much more explicit, and
> >>>> provide the reason why.
> >>>>
> >>>> How about:
> >>>>     cc: <do-not-apply-to-stable@xxxxxxxxxx> # Reason goes here, and must be present
> >>>>
> >>>> and we can make that address be routed to /dev/null just like
> >>>> <stable@xxxxxxxxxx> is?
> >>>
> >>> Totally fine with me, but that feels somewhat long and hard to type.
> >>
> >> I want it long and hard to type and very very explicit that this is what
> >> the developer/maintainer wants to have happen (again, because this is
> >> such a rare occurrence.)
> >>
> >>> How
> >>> about just 'no-stable@xxxxxxxxxx' (or 'nostable@xxxxxxxxxx')?
> >>
> >> More words are better :)
> > 
> > And after that, someone discovers this turns out to be (a hard
> > dependency for) a very critical fix that does need backporting?
> 
> Ask why the tag was set I guess. But yeah, that was among the minor
> reasons why I had come up with "no semiautomatic stable backport" thing,
> as it made the intention more clear. Maybe
> 
> only-manual-stable-backport@xxxxxxxxxx
> 
> could help and is even longer. But I might be getting into bikeshedding
> territory here. :-D

That one would not work as I would then manually backport the commit :)

Actually, one can say that all of the commits are manually backported as
I review them all that are cc: stable when I apply them.  So while
bikeshedding is fun, this would mean the opposite of what you intend.

thanks,

greg k-h